Quantum-computing startup Rigetti to offer modular processors

One quantum-computing startup said Tuesday that it will be making new designs for its quantum processors. Rigetti Computing won't build a monolithic processor like everyone else. Instead, Rigetti Computing will create smaller collections of qubits on chip that can physically be linked together to form a single functional processor. This is not multiprocessing, but modular chip design.This is a significant move for Rigetti processors as well as quantum computing in general.What's the problem?Rigetti's computers are based on technology known as a "transmon" which is based upon a superconducting wire loop connected to a resonator. It is the same qubit technology that large rivals like Google and IBM use. Quantum computing is unique in that the state of one transmon can have an effect on its neighbor's during calculations. The topology of connections between transmon qubits can be a significant contributor to the machine’s computational power.This is in contrast with Honeywell's Ion-trap computer in which any qubit can interact to any other qubit at most at the current qubit count.Performance is also affected by the error rate for individual qubits as well as the number of qubits. The ability to increase the number of qubits can improve the processor's computational power, but only if the quality of all qubits is sufficient that the error rate does not limit the processor's ability to do accurate computations.When qubit counts exceed the thousands, error correction is possible which significantly changes the process. We currently have less than 100 qubits. This change will be made in the near future.What has changed for Rigetti?Rigetti is a startup and doesn't have the same resources as companies like IBM. Rigetti has been able to produce its own transmon processing units, but it still falls behind larger companies. Rigetti's current processors have a qubit count of 31. IBM's new processors boast a more than 60 qubits.AdvertisementThis has not been a problem at a time where we are still uncertain if quantum processors can perform useful calculations without sufficient qubits for error correction. However, it is crucial to have a clear path for rapid scale in order to reach error correction. We may discover that certain algorithms can be successfully run on qubits between the ones currently available and those required for full error correction.Rigetti believes that the ability to combine several smaller processors, which the company has demonstrated it can produce, into one larger processor will allow it to quickly increase its qubit count. Today's announcement by IBM stated that it expects to have an 80-qubit processor available in the next few months. (To give context, IBM's roadmap also includes plans to release a 127-qubit processor this year.Moving away from monolithic designs has another advantage. Most chips have one or more qubits with a high error rate or are defective. Modular design can help reduce these consequences. Rigetti is able to manufacture a large number of modules and assemble chips with the fewest defects.Alternativly, the company might choose modules with low error rates that have qubits and create the equivalent of an all star processor. A lower qubit count could have an offset effect.The larger pictureQubit technologies, as it stands now, require components that can be larger than any component in a computer chips. Transmons and optical quantum computers both require cabling (optical compatible or microwave compatible), while atoms in ion trap computers are held in place by an array electrodes. Scaling the number of qubits to match traditional silicon chips is not as easy as adding computational resources. The physical constraints are too different.Other quantum-computing companies that we have spoken to already acknowledge the need to find ways to incorporate qubits on multiple chips. Rigetti may have found a solution to a problem other companies will likely face.Transmons, however, can be connected by wiring and are expected to be one the easiest technologies to use in this regard. These lessons won't be applicable to other technologies such as trapped ions.